OpenAI has opened access to fine-tuning the GPT-3.5 Turbo language model

OpenAI has allowed business customers to refine the GPT-3.5 Turbo language model using their own data to solve certain problems.

Here's What We Know

Tweaking allows ChatGPT to be "honed" to a more focused model that is particularly effective at solving certain tasks. Guided training can result in a unique bot that responds, for example, in a specific language or in more concise terms.

GPT-3.5 Turbo has knowledge until September 2021, but businesses will be able to provide it with additional data. OpenAI says it will not use user data to improve the base model.

Flashback

GPT-3.5 Turbo is a family of models that OpenAI introduced in early 2023. According to the company, it is ideal for use cases not related to chat.

The model is capable of handling 4,000 tokens at a time, which is double what its predecessors supported. The company also added that the thin instructions allowed the first testers to make the prompts 90 per cent shorter.

GPT-3.5 costs $0.0080 per 1,000 tokens for training, $0.0120 per 1,000 tokens for input usage, and $0.0120 per 1,000 tokens for chatbot output.

Source: The Verge